I have two groups, in which three possible outcomes appeared at different frequencies, and I wanna know, if there is a significant difference in-between the groups.

For instance, I have two potatoe fields, field A and field B. The potatoes grew either tall, normal, or small in the respected fields e.g.:

    Amount of potatoes
    Tall  -   Normal  -  Small
A:   30         10       2
B:   48         12       1

"71% percent of the potatoes on field A grew tall." Is this significantly different to on field B?

Was looking for some Anova test but couldn't find a suitable. Also most were looking for numeric data with mean values etc.

  • $\begingroup$ A chi square test is probably what you are looking for. This is a better question for the cross validate site. $\endgroup$
    – Dave2e
    Commented Apr 29, 2022 at 22:57
  • $\begingroup$ Thanks Dave! Chi square looks good. But it is only usable for two variables, is that correct? $\endgroup$
    – Roberto
    Commented Apr 29, 2022 at 23:23
  • 1
    $\begingroup$ Your example only has two variables. The loglin() function provides a way to handle multidimensional tables. Also a possibility is Poisson regression using glm(). $\endgroup$
    – dcarlson
    Commented Apr 30, 2022 at 3:57
  • 1
    $\begingroup$ $\chi^2$ is good for any number of variables. I also like the $G-test$, which is a likelihood ratio test of nested (multinomial) logistic regression models. I show that here, and you can apply my code to nested multinomial logistic regression models, since you have a $3$-category $y$. $\endgroup$
    – Dave
    Commented Apr 30, 2022 at 9:28


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.